Researchers at MIT have developed a method that improves the accuracy of predictions generated by climate models. The technique involves the use of machine learning and dynamical systems theory to make predictions from coarse climate models more accurate. These models, which are used to predict the impact of climate change including extreme weather events, work on a large scale and can often miss important details at smaller scales such as individual cities. The new method acts as a correction scheme, allowing the model’s simulations to provide more accurate information.
The team’s approach involves an algorithm that is layered over the model’s output. This algorithm nudges the output closer to real-world conditions and learns associations within data such as past temperature and humidity, to correct the model’s predictions. Applying this method, the team found that corrected models produced climate patterns that closely matched real-world observations over a 36 year period.
Differences in predictions between uncorrected and corrected models might seem small, but can have a significant impact on humans experiencing such events. For example, an uncorrected model might predict an extreme weather event with a temperature of 105°F, while the corrected model might predict the same event with a temperature of 115°F. This difference in temperature can significantly impact individuals experiencing such an event and have broader implications for preparation and response to extreme weather events.
The method is also able to be applied to any global climate model. This has the potential to provide more accurate understanding of where and how often extreme weather events will occur in the future. This can assist in preparation as well as the development of solutions to cope with the effects of climate change. Future work on this project will focus on analyzing future climate scenarios.
Overall, this new method allows for greater accuracy in predicting the frequency of extreme weather events, which is vital for preparing for the impacts of climate change. Not only is this a novel approach to climate modeling, it also has potential practical applications in the face of our changing climate. This research was supported in part by the U.S. Department of Energy, and the results were published in the Journal of Advances in Modeling Earth Systems.
